ChatCTNNB1

CTNNB1 Connect and Cure, Inc.

There are approximately 30 million rare disease patients in the United States, at least 50% of which are undiagnosed. To date, over 7,000 rare diseases have been identified (1) and FDA-approved treatments exist only for about 5% of these conditions. Patients not being diagnosed poses a problem for disease communities with and without approved treatments alike. 

A patient suffering from a disease with an approved treatment cannot be treated until they’re diagnosed.

Simultaneously, treatments will never be developed for diseases for which too few patients are diagnosed. Large groups of diagnosed patients are required to curate medical knowledge about the condition, to recruit patients into clinical trials, and to convince drug developers of the potential return on investment of developing a treatment for a given disease.
There are fewer than 500 cases of CTNNB1 syndrome reported globally. However, it is estimated that 3 individuals in every 100,000 births may be affected with CTNNB1 syndrome (2). Patients with disease-causing variants in the CTNNB1 gene experience complex neurological symptoms, including cognitive impairment, developmental delays, and skeletal abnormalities (2). CTNNB1 Connect and Cure seeks to advance the development of an artificial intelligence (AI) platform to identify undiagnosed CTNNB1 patients and spur drug development efforts for CTNNB1 syndrome. The AI platform us ubiquitously applicable to most of the over 7,000 rare diseases.

1. https://rarediseases.org/wp-co...
2. https://rarediseases.org/rare-...

The proposed solution uses AI to identify undiagnosed CTNNB1 patients and provides them with the appropriate genetic testing.

Patients or their caretakers land on a website, have a conversation with a large language model, and upload a facial photo of the patient. The AI algorithms analyze the data and predict if the patient has undiagnosed CTNNB1 syndrome. If the patient is predicted to have CTNNB1 syndrome, the patient is tested using whole genome sequencing via an at-home telemedicine-enabled clinical genetic testing service. All data collected on CTNNB1 patients will be used to advance drug development for CTNNB1 syndrome.

The platform combines innovations in large language models (OpenAI 2023), facial recognition (Nature Genetics 54, 349–357 (2022)), and bioinformatics.

The currently-live version of the platform can be viewed at chat.probablygenetic.com.

The proposed solution is a further advancement of a technology developed by Probably Genetic Inc. Probably Genetic’s original technology platform uses different AI techniques, not including large language models and facial recognition, to identify undiagnosed patients. CTNNB1 Connect and Cure, Inc. has previously partnered with Probably Genetic and successfully identified undiagnosed CTNNB1 patients using their technology. 

Our solution will serve diagnosed CTNNB1 patients, undiagnosed CTNNB1 patients, physicians treating rare neurological disease patients, and drug developers. Our diagnosed CTNNB1 patient population will participate in generating a comprehensive database of the symptoms and timeline of symptom onset experienced by CTNNB1 syndrome. They will utilize Probably Genetic’s patient-friendly data collection methods to report symptoms in an unrestricted format. The collected data will be utilized to generate novel literature that better describes the symptoms associated with CTNNB1, which may assist physicians with reducing the diagnostic odyssey and support drug development efforts. Additionally, the data will be used to train a large language model to identify potential CTNNB1-positive, undiagnosed patients. Undiagnosed patients will benefit from whole genome sequencing to confirm their diagnosis. This solution could end a years long diagnostic odyssey or incorrect differential diagnosis that most rare disease patients experience. Additionally, Probably Genetics' novel AI platform can be trained on all rare diseases, thus drastically reducing the diagnostic journey for all rare disease patients. 

Probably Genetic’s novel platform combining large language models and facial recognition is currently being evaluated in over 10 disease areas with multiple patient advocacy group and drug development partners. Improvements to the underlying AI system and the platform’s user experience are being implemented. 

Probably Genetic’s novel AI platform that we propose to assess for CTNNB1 patient identification has been used by several thousand patients. Probably Genetic’s predecessor to this novel platform has served over 100,000 patients to date.

Probably Genetic’s AI platform has the potential to reduce the diagnostic odyssey for all rare diseases. 

If all undiagnosed patients had instantaneous access to a free physician who perfectly knows which patients to test, the diagnostic odyssey wouldn’t exist. Patients would get diagnosed immediately, and payers would reimburse every single test ordered by that doctor because the cost savings would be obvious. Most importantly, drug developers would compete to treat these patients because all patients are identified. Probably Genetic’s system can become this instantaneously accessible free doctor with perfect selection for testing. Their system could live in various distribution channels and reach patients anywhere.

To make Probably Genetic’s system perfect, it needs to have a high precision P:

P = TP/(TP+FP),

where TP = true positives, i.e. number of patients the system thought had a specific disease that has that specific disease, and FP = false positives, i.e. number of patients the system thought had a specific disease that doesn't have that specific disease.

Diagnosing all US rare disease patients without their system would cost $330B (assuming a US population of 330M and a cost per WGS test of $1,000).

For simplicity’s sake, assume we had a system with 100% Precision. To diagnose every rare disease patient in the US, we wouldn’t need $330B. If we assume that 1 in 10 people in the US have a rare disease, we’d only want to test 33M people for a total cost of $33B, 10x cheaper than testing everyone. If we got even better and knew how to select not just rare disease patients, but the 40% of rare disease patients for which WGS can deliver a diagnostic result today, we could reduce the total cost to diagnose all diagnosable US rare disease patients to $13.2B, a small number compared to Everylife’s estimate that rare diseases cost the US $1 trillion in 2019.

A system with 10% precision would also make patients with any specific rare disease 10x cheaper to find. Many diseases would become attractive business opportunities for drug developers. 

For example, imagine trying to find a patient with a disease with a prevalence of 1/100,000. At $1,000 per test, you could spend up to $100,000,000 to find one patient if you randomly sampled the population. However, if you had a precision of just 10%, you’d only have to test 10 people to find one patient with the target disease and thus lower your cost per diagnosis to $10,000. Pharma companies are willing to spend 10%-30% of the revenue they make per patient in the first year of treatment just to identify the patient. That would mean that it may all of a sudden be financially attractive to develop treatments that “only” get reimbursed at $10,000/30% = $33,333. That’s approximately 10x lower than the >$300,000 most pharma companies are currently targeting for treatment prices.

High-precision systems can track clinical outcomes effectively, holding drug developers accountable for patient improvement.

We are applying for The Amgen Prize to assist our partnership with Probably Genetic in funding and incentivizing already-diagnosed CTNNB1 patients in gathering a comprehensive database of the CTNNB1 phenotype, training their large language model, ChatCTNNB1, and funding free whole genome sequencing necessary for validation of their AI platform.

Our team is well-positioned to deliver this solution as we have a large enough patient population to submit training data to Probably Genetics platform, as well as an outreach team ready to engage with our diagnosed families, as well as undiagnosed communities to help locate potential undiagnosed CTNNB1 patients. Many of our own families remain active in the community channels they originally surfaced on when they were looking for answers to the symptoms their children were experiencing. 

Additionally, Probably Genetic has a proven track record in rare disease machine learning development and is currently running this approach in a number of disease areas. They have already been working closely with our community to gather feedback on their platform to optimize its user experience and ability to capture and understand the CTNNB1 patient experience. 

Our short term goals include enrollment for our Natural History Study, Patient Registry, and Biobank efforts all of which are crucial to making progress in our treatment development progress. We hope to be Phase I clinical trial-ready in the next two years, which is directly tied to the success of the goals mentioned previously, all of which will be more successful by increasing the number of known CTNNB1 Syndrome patients. We have ongoing efforts  in increasing awareness of the justification for ordering genetic testing for patients diagnosed with cerebral palsy, as it is a common misdiagnosis for CTNNB1 Syndrome, as well as hope to continue and expand upon our work with Probably Genetic to diagnose CTNNB1 patients. 

Our solution will increase the current phenotypic understanding of CTNNB1 syndrome, as well as other rare diseases, give patients access to genetic sequencing, and shorten the rare disease diagnostic odyssey. We will increase the current phenotypic understanding of CTNNB1 syndrome by generating a comprehensive database of the CTNNB1 phenotype and timeline for symptom onset through a partnership between CTNNB1 Connect and Cure, a patient advocacy group, and Probably Genetic, a rare disease company employing the use novel artificial intelligence and large language models with patient-friendly user settings. Clinically diagnosed CTNNB1 patients’ phenotypic data will be used to train a large language model to detect undiagnosed CTNNB1 patients, who are then validated through free access to whole genome sequencing to detect disease-causing variants to the CTNNB1 gene. Additionally, the comprehensive data collected can be utilized by pharmaceutical companies to determine novel treatment targets or cures for CTNNB1 syndrome. Through our solution, diagnosed CTNNB1 patients can assist us in identifying undiagnosed CTNNB1 patients using their understanding of this rare disease, they will help to bolster the current literature on CTNNB1, and aid physicians in better identifying potential CTNNB1 patients in clinics across the world. Undiagnosed individuals benefit from an end to a potentially years-long diagnostic odyssey and misdiagnoses. It is critical to understand that the project described by our solution can be applied to any number of rare diseases – with the ability to generate comprehensive databases for all rare diseases, and ending the diagnostic journey for millions of patients.

We have a twelve-person parent advisory board and a person science advisory board. This grant would also look to engage a technical and operational team of 20 of Probably Genetic’s team members to train the model and fund testing. 

Probably Genetic started developing their technology platform in 2019. 

The CTNNB1 Connect and Cure board of directors consists of 4 males and 5 females, all parents to different aged individuals with CTNNB1 Syndrome. Multiple countries, races, burden of disease, and socioeconomic statuses are represented in our board. Its governance structure not only ensures diversity in decision-making processes, but it also strengthens our organization's access to the members of our community. We are actively working on ways to grow our diversity further by collaborating with other CTNNB1 groups in other countries, and forming a Parent Advisory Board. CTNNB1 Connect and Cure is also focused on ensuring equity and inclusion. Some other projects underway are regular virtual community meetings and translating communications and materials into other languages. We are confident to bring our core value of diversity and inclusion to our project, as we will be working to raise awareness about the program in online channels, and focus on traditionally underserved communities. 

CTNNB1 Connect and Cure was established as a 501c3 nonprofit, registered in Delaware, to find treatments for people with CTNNB1 Syndrome and build community among affected families. CTNNB1 Syndrome is a rare neurodevelopmental disease with no approved treatments yet. The CTNNB1 gene is needed in virtually all cells of the body, so the effects of mutations of this gene are significant. People with CTNNB1 Syndrome have difficulty in all areas of life, including the ability to walk, talk, and think. Several severe conditions are associated with the disease such as retinopathy that can lead to blindness, tethered cord that can lead to nerve damage, and progressive spasticity that can lead to contractures. Some patients experience epilepsy or congenital heart defects. CTNNB1 syndrome presents a vast array of difficulties for the diagnosed individual and their entire family. With fewer than 500 patients known worldwide, it is up to our organization to fund the research and facilitate the connections needed to develop treatments.

Our business model overview is to progress the development of treatments and connect with families by actively fundraising and applying for grants. Most importantly, we are 100% volunteer-led so that our money can be spent directly towards our mission. This requires active recruitment of volunteers who are experienced in the areas of our operations, which fall into three main categories: internal, operations, and external. Our executive team is made of a CEO, a CFO, a COO, and a CCO, and there are 9 voting directors. Since our patient population is spread out all over the world, we place focus on our virtual communications such as our website, social media, and email subscription lists. Our board of directors is similarly spread out, so our meetings and day-to-day interactions are also virtual. Most of the incoming funds are from parent-led fundraising campaigns, so we provide support in their efforts by creating premade resources and using a user-friendly fundraising platform. CTNNB1 Connect and Cure works closely with its Scientific Advisory Board, made of experienced clinicians and researchers, to make sure we are taking the correct steps in the correct order, all so that we can find the best treatments for as many CTNNB1 patients as efficiently as possible. So far, our organization has funded a mouse model, preclinical testing of a novel small molecule treatment in animal models and patient-derived cells, hosted two virtual conferences and one in-person conference, a biomarker study, and in-person clinical evaluations for a natural history study. We actively collaborate with other CTNNB1 organizations around the world as well as organizations of other rare diseases.

We work to consistently identify grant funding opportunities, as well as run fundraising drives within our community. We keep our expenses low by running our board through volunteered time, and focus on sourcing non-restricted donations. Additionally, we sell merchandise through the website, and will come out with specialty lines around important rare disease events, such as Rare Disease Day.